1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,prb]{revtex4}
2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3:
4: % Some other (several out of many) possibilities
5: %\documentclass[preprint,aps]{revtex4}
6: %\documentclass[preprint,aps,draft]{revtex4}
7: %\documentclass[prb]{revtex4}% Physical Review B
8:
9: \usepackage{graphicx}% Include figure files
10: \usepackage{dcolumn}% Align table columns on decimal point
11: \usepackage{bm}% bold math
12: \begin{document}
13:
14: %\preprint{APS/JCP-IS}
15:
16:
17: \title{Taming the Rugged Landscape: Techniques for the Production,
18: Reordering, and Stabilization of Selected Cluster Inherent Structures }
19:
20:
21: \author{Dubravko Sabo}
22: \author{J. D. Doll}
23: \affiliation{Department of Chemistry, \\ Brown University, \\ Providence,
24: RI 02912, USA}
25:
26: \author{David L. Freeman}
27: \affiliation{Department of Chemistry, \\ University of Rhode Island,
28: \\ Kingston, RI 02881, USA}
29:
30:
31: \date{\today}
32:
33:
34: \begin{abstract}
35: We report our studies of the potential energy surface (PES) of selected binary
36: Lennard-Jones clusters. The effect of adding selected impurity atoms
37: to a homogeneous cluster is explored. Inherent structures and transition
38: states are found by combination of conjugate gradient and
39: eigenvector-following methods while the topography of the PES is mapped
40: with the help of a disconnectivity analysis. We show that we can controllably
41: induce new structures as well as reorder and stabilize existing structures
42: that are characteristic of higher-lying minima.
43: \end{abstract}
44:
45:
46: % insert suggested PACS numbers in braces on next line
47: \pacs{82.20.Wt,02.60.Pn}
48:
49: \maketitle
50:
51: \section{Introduction} \label{sec:intro}
52:
53: The minimization/optimization problem is one of the more ubiquitous and
54: challenging in computational science \cite{PRESTEU}.
55: Central to researchers in the physical
56: sciences and engineering, this problem is also of primary importance to
57: social, biological, and economics investigators.
58:
59: Driven in large measure by such widespread interest, there has been appreciable
60: progress on the minimization problem. Especially notable have been algorithmic
61: advances in the form of annealing and stochastic relaxation approaches
62: \cite{PRESTEU,KIRGEVE,SCHERAGA86,SCHERAGA87,FAKEN99,BERNE00}
63: as well as basin-hopping techniques \cite{WALES97B,WALES99S}.
64: In both classical and quantum form, these methods
65: offer valuable, complementary alternatives to traditional, gradient or
66: pseudo-gradient approaches \cite{PRESTEU}.
67:
68: In addition to algorithmic developments relevant to the minimization problem,
69: there have also been notable advances in the tools to classify and analyze
70: the topography of the underlying objective functions. In chemical applications,
71: the principal focus of the remainder of our discussion, the objective
72: function of interest is typically a specified potential or free energy surface.
73: Following Stillinger and Weber \cite{STILLINGER83,STILLINGER84},
74: it is useful to perform
75: an ``inherent structure'' decomposition of the associated configuration space
76: by employing the minima (local and global) of this surface. These inherent
77: structures, their relative orderings, and their connectivity provide
78: important information concerning the structure, function, and dynamics of
79: the associated physical system. Disconnectivity analysis introduced by
80: Czerminski and Elber \cite{ELBER90}, discussed by
81: Becker and Karplus \cite{KARPLUS97} and developed by,
82: among others, Wales, Doye and Miller \cite{WALES99A,WALES99B,WALES99C}
83: has proved especially valuable with respect to these latter tasks.
84:
85: As evidenced by the development of classical and quantum annealing methods,
86: there is an important interplay between minimization and the Monte Carlo
87: sampling problem. Both applications, for example, are concerned with
88: overcoming barriers that inhibit the interconversion or isomerization
89: of the various inherent structures of the problem. Consequently,
90: developments in one field contain implications for developments in the
91: other. Advances in rare event sampling methods, such as
92: J-walking \cite{FREEMAN92,JORD93C} and
93: parallel tempering \cite{PARISI92,THOMPSON95,FREEMAN00A,FREEMAN00B}
94: methods, thus contain implications for the minimization problem.
95:
96: The field of atomic and molecular clusters has been and continues to be
97: an important test bed for the development and application of minimization
98: and analysis methods. Utilizing the methods outlined above,
99: researchers have produced a relatively coherent picture of the relationship
100: between the nature of underlying PES and the physical properties of the
101: associated systems. For example, from the studies by Berry {\it et al.}
102: \cite{BERRY86,BERRY90,BERRY93,BERRY94},
103: the single component studies of Wales {\it et al.}
104: \cite{WALES99A,WALES99B,WALES99C},
105: the mixed LJ cluster studies of Jordan {\it et al.} \cite{JORD02}
106: as well as the research of others \cite{ABRAHAM78,FRANTZ96,FRANTZ97,
107: SERRA97},
108: we have begun to understand the nature of systems for which the lowest
109: inherent structures can or cannot be readily located.
110:
111: In present paper we would like to build upon advances in the minimization
112: problem by effectively turning the logic ``upside down''. That is,
113: instead of asking what we have to do in order to locate or sample
114: the global minimum of a specified potential energy surface, we wish
115: instead to ask how we might go about controllably inducing new structures
116: as well as reordering and stabilizing existing structures that are
117: characteristic of higher-lying local minima. Basically, we seek to utilize
118: what we have learned about what it takes to {\em avoid} local minima
119: to instead {\em controllably produce} them.
120:
121: In principle, one can envision efforts involving both thermodynamic and
122: kinetic approaches. In the present work we shall focus principally on
123: the thermodynamic issues. Furthermore, we shall limit the discussion
124: in the present work to applications involving clusters. As discussed
125: elsewhere \cite{JELL99},
126: clusters are of appreciable technological importance,
127: are valuable as prototypes for the study of the properties of extended
128: systems, illuminate issues related to the size-dependence of selected
129: physical properties, and provide valuable test beds for the development
130: and application of emerging computational techniques. This combination
131: of formal, computational and technological interest has produced a
132: vast and growing cluster literature \cite{FREEMAN96,JELL99,KNICKEL99}.
133:
134: The remainder of the paper is organized as follows:
135: In Section~\ref{sec:comput} we outline the computational details
136: of the present study. We discuss the methods we use to determine
137: the inherent structures and transition states of a specified
138: cluster's potential energy surface. Using these methods, we examine
139: specific results for two prototype systems in Section~\ref{sec:numres}.
140: These particular results are
141: designed to demonstrate ``proof of principle'' with respect to
142: the basic objectives of the present study for selected systems.
143: Finally, in Section~\ref{sec:conclude} we summarize our results
144: and speculate about likely future research
145: directions.
146:
147:
148: \section{Computational Details} \label{sec:comput}
149:
150: The present Section describes the computational details of our
151: investigations involving binary clusters of the form $\mathrm{X_nY_m}$. Our
152: overall interest will be to explore the extent to which we can
153: utilize the ``adatoms'' (i.e. the Y-system) to induce, reorder and
154: stabilize selected inherent structures in the ``core'' X-system. While
155: one can easily imagine applications involving both more and more
156: complex components, we feel these relatively simple, two-component
157: clusters are a convenient starting point for an initial study of the
158: issues we raise.
159:
160: We shall assume in what follows that the total potential energy is
161: composed of a pairwise sum of Lennard-Jones interactions. Specifically,
162: we assume that the total potential energy, $\mathrm V_{tot}$, for an
163: N-particle system is given by
164:
165: \begin{equation}
166: \label{2.1}
167: V_{tot} = \sum_{i<j}^{N} v_{ij}(r_{ij}),
168: \end{equation}
169: where the pair interaction as a function of the distance between
170: particles i and j, $r_{ij}$, is given by
171:
172: \begin{equation}
173: \label{2.2}
174: v_{ij}(r_{ij}) = 4\epsilon_{ij}\
175: [( \frac{\sigma_{ij}}{r_{ij}})^{12}
176: -( \frac{\sigma_{ij}}{r_{ij}})^{6}] .
177: \end{equation}
178: In Eq. (\ref{2.2}) the constants $\epsilon_{ij}$ and $\sigma_{ij}$ are
179: the energy and length-scale parameters for the interaction of
180: particles i and j.
181:
182: For a two-component system, we must specify both the ``like''
183: (X-X, Y-Y) as well as the ``mixed'' (X-Y) interactions. With an eye
184: toward studying trends in the results as opposed to results for
185: particular physical systems, it is convenient to reduce the number
186: of free parameters. To do so, we shall assume in the present study
187: that the ``mixed'' Lennard-Jones values are obtained from the ``like''
188: Lennard-Jones parameters via usual combination rules \cite{COMBINE}
189:
190: \begin{equation}
191: \label{2.3}
192: \sigma_{_{XY}}= \frac{1}{2}(\sigma_{_{XX}}+\sigma_{_{YY}})
193: \end{equation}
194:
195: \begin{equation}
196: \label{2.4}
197: \epsilon_{_{XY}}= \sqrt{\epsilon_{_{XX}}\epsilon_{_{YY}}} .
198: \end{equation}
199: Furthermore, we note that with the mixed Lennard-Jones parameters
200: specified as in Eqs.(\ref{2.3}) and (\ref{2.4}), the inherent structure
201: topography of the ``reduced'' potential energy surface of the binary
202: system (i.e. $V_{tot}/\epsilon_{_{XX}}$) is a function of only two
203: parameters, ($\sigma, \epsilon$), the ratios of the corresponding
204: adatom/core length and energy parameters
205:
206: \begin{equation}
207: \label{2.5}
208: \sigma= \sigma_{_{YY}}/\sigma_{_{XX}}
209: \end{equation}
210:
211: \begin{equation}
212: \label{2.6}
213: \epsilon= \epsilon_{_{YY}}/\epsilon_{_{XX}} .
214: \end{equation}
215:
216: If necessary for a discussion of a specific physical system, the
217: absolute bond lengths, energies, activation energies, etc. can be
218: obtained from the corresponding ``reduced'' results by a simple
219: rescaling with the appropriate core-system Lennard-Jones parameters.
220:
221: The computational task in our study is thus one of exploring and
222: characterizing the (reduced) potential energy surface of our
223: binary cluster systems as a function of the number of (core, adatom)
224: particles, (n,m), and for given ($\sigma,\epsilon$) ratios. In
225: typical applications the lowest $\mathrm N_{IS}$ inherent structures and the
226: associated disconnectivity graphs are determined. For the applications
227: reported here, $\mathrm N_{IS}$ is generally of the order of a few hundred
228: (thousand) or less. Depending on the size of the cluster, inherent structures
229: are found either via conjugate gradient methods starting from randomly
230: chosen initial configurations, or by more systematic surface exploration
231: methods such as those outlined by Wales and co-workers \cite{WALES99B}
232: and by Jordan {\it et al.} \cite{JORD93B}.
233: In all cases, the inherent structures that are located are
234: confirmed to be stable minima via a standard Hessian analysis. To
235: reduce the chance we miss particular local or global minima, we monitor
236: the number of times individual inherent structures are found and
237: demand that each of the N$_{IS}$ inherent structures be located a
238: minimum number of times (at least 10) before we terminate our search.
239: Once we are satisfied we have located the relevant inherent structures,
240: transitions states linking these stable minima are obtained using the
241: eigenvector following methods outlined by Miller and Cerjan \cite{MILLER81}
242: and further developed by Simons {\it et al.} \cite{SIMONS83,SIMONS85,SIMONS90},
243: Jordan {\it et al.} \cite{JORD93B} and Wales \cite{WALES94}.
244: Finally, with the requisite inherent structures and
245: barriers in hand, we perform a disconnectivity analysis using methods
246: outlined by Czerminski and Elber \cite{ELBER90},
247: Becker and Karplus \cite{KARPLUS97} and Wales {\it et al.}
248: \cite{WALES99A} .
249:
250:
251:
252: \section{Numerical Results} \label{sec:numres}
253:
254: In the present Section, we wish to illustrate the general themes we
255: introduced in Section~\ref{sec:intro} .
256: We do so by demonstrating that we can accomplish
257: three basic objectives. Specifically, we show that by adding selected
258: ``impurity'' atoms to bare ``core'' systems, we can:
259:
260: 1. induce new ``core structures''
261:
262: 2. reorder the energies of existing core inherent structures, and
263:
264: 3. stabilize selected inherent structures by controlling the activation
265: energies that determine their isomerization kinetics.
266:
267: For purposes of illustration, we shall examine numerical results for a few,
268: simple Lennard-Jones systems involving five and seven core atoms, systems
269: well-known from previous studies to have one and four energetically distinct
270: inherent structures, respectively. The inherent structures and their
271: associated energies for these core systems are illustrated in Fig.
272: ~\ref{fig:core5} and ~\ref{fig:core7}.
273: \begin{figure}[!tbp]
274: \includegraphics[clip=true,width=3.2cm]{fig_1.ps}
275: \caption{\label{fig:core5} The only stable inherent structure for X$_5$
276: LJ cluster. Its energy (in units of the LJ well depth) is -9.104.}
277: %Its energy (in units of the LJ well depth) is -9.10385242.
278: \end{figure}
279:
280: \begin{figure}[!htbp] \centering
281: \begin{tabular}{@{}cc@{}}
282: \includegraphics[width=4.2cm,clip=true]{min1.ps} &
283: \includegraphics[width=4.2cm,clip=true]{min2.ps} \\
284: (a) & (b) \\
285: \includegraphics[width=4.2cm,clip=true]{min3.ps} &
286: \includegraphics[width=4.2cm,clip=true]{min4.ps} \\
287: (c) & (d) \\
288: \end{tabular}
289: \caption{\label{fig:core7}
290: The four, energetically distinct, stable inherent structures for
291: X$_7$ LJ cluster. The energies (in units of the LJ well depth) are: (a)
292: -16.505, (b) -15.935, (c) -15.593, (d) -15.533.}
293: %(a)-16.50538417, (b) -15.93504306, (c) -15.59321094, (d) -15.53306005.}
294: \end{figure}
295:
296: We first consider mixed clusters of the generic type X$_5$Y$_2$. Here two
297: impurity Y-atoms are added to the parent, five-atom X-core. We have chosen
298: this system because it builds upon the very simple five-atom core, a system
299: that has only a single inherent structure, and because the total system has
300: a total of seven atoms, a magic number for icosahedral growth in homogeneous
301: systems. Using the techniques of Section II, we then determine the lowest
302: several inherent structures for a range of ($\sigma$,$\epsilon$) [c.f.
303: Eq.(\ref{2.5}) and Eq.(\ref{2.6})].
304: As can be seen from Fig.~\ref{fig:totPES52}, the total
305: potential energy [Eq.(\ref{2.1})] of the lowest inherent structure for the
306: X$_5$Y$_2$ system shows no appreciable structure as a function of the
307: ($\sigma$,$\epsilon$) parameters.
308: \begin{figure}
309: \includegraphics[clip=true,width=8.5cm]{fig_3.font.ps}
310: \caption{\label{fig:totPES52}
311: $\mathrm E_{tot}(\sigma,\epsilon)$ (c.f.Eqs.\ref{2.1}, \ref{2.5} and \ref{2.6})
312: for the $\mathrm{X_5Y_2}$ system. Note the relative lack of structure in the
313: ($\sigma$,$\epsilon$) variation of the total cluster energy.}
314: \end{figure}
315:
316: On the other hand, we see in Fig.~\ref{fig:corePES52}
317: that the core potential energy, defined as the
318: potential energy of interaction for only the core X-atoms, of the minimum
319: (total) energy cluster clearly breaks into extended regions, each
320: corresponding to a well-defined core structure. The reader should notice
321: that each region in Fig.~\ref{fig:corePES52} contains the same ``kind''
322: of core structure but their core energies are slightly different. We have
323: chosen a single ``average'' core energy value to represent all energies in the
324: corresponding domain for plotting convenience.
325:
326: The distinct core structures, shown in Fig.~\ref{fig:corePES52}, have
327: been identified by examining their core energies (E$_{core}$) and their
328: principal moments of inertia. For each structure a triplet of
329: values (E$_{core}$, I$_2$, I$_3$) has been associated, where I$_2$ and
330: I$_3$ are the moments of inertia about the principal axes 2 and 3,
331: respectively. We have defined I$_2$ and I$_3$ in the following way:
332: I$_2$=I$^{'}_2$/I$^{'}_1$, I$_3$=I$^{'}_3$/I$^{'}_1$ where I$^{'}_1$,
333: I$^{'}_2$ and I$^{'}_3$ are the principal moments of inertia obtained
334: by diagonalizing the inertia tensor of the system. If the triplet of
335: values has not been sufficient to identify a core structure then we
336: have examined the structure visually.
337:
338: \begin{figure}[!htbp]\centering
339: \includegraphics[clip=true,width=8.5cm]{fig_4.font.ps}
340: \caption{\label{fig:corePES52}
341: $\mathrm E_{core}(\sigma,\epsilon)$ for the $\mathrm{X_5Y_2}$ system.
342: Here the ``core'' energy is defined as that portion of the potential energy
343: arising from only the core-core atom interactions. Unlike the total energy,
344: the $\mathrm(\sigma,\epsilon)$ variation of the core cluster energy exhibits
345: relatively well-defined regions. The labels of each of these regions in
346: the figure correspond to the distinct core structures shown in
347: Fig.~\ref{fig:coreE52}.}
348: \end{figure}
349:
350: Selected cluster structures
351: illustrating the core arrangements corresponding to various
352: ($\sigma$,$\epsilon$) values are shown in Fig.~\ref{fig:coreE52}. We see from
353: Figs.~\ref{fig:corePES52} and ~\ref{fig:coreE52} that the $\mathrm{X_5Y_2}$
354: cluster exhibits core X-atom structures that include trigonal bipyramidal,
355: planar and square pyramidal core geometries. Of these, only the trigonal
356: bipyramidal form is stable in the parent $\mathrm X_5$ system. This
357: illustrates that a suitable choice of the ($\sigma$,$\epsilon$) parameters
358: can controllably induce core geometries not present as stable minima in
359: the bare cluster. For example, the square pyramid core structure, seen
360: in Fig.~\ref{fig:coreE52}.2 as a stable system, corresponds to a transition
361: state in the bare $\mathrm X_5$ cluster.
362:
363: Fig.~\ref{fig:DisCon52} represents the $\mathrm{X_5Y_2}$ cluster at
364: four points in Fig.~\ref{fig:corePES52} defined by the ($\sigma$,$\epsilon$)
365: coordinates (0.4,0.5), (0.4,1.0), (0.4,1.5) and (0.4,2.0).
366: Here the pairs of coordinates correspond
367: to (a), (b), (c) and (d) part of Fig.~\ref{fig:DisCon52}, respectively.
368: In other words, we keep value of $\sigma$=0.4 fixed, while increasing
369: the value of $\epsilon$.
370: Each disconnectivity graph shows all inherent structures available to
371: the system for the given ($\sigma$,$\epsilon$) values. The global minimum
372: of each system is labeled by number 1 and contains as a recognizable
373: component the square pyramid core structure (see Fig.~\ref{fig:coreE52}.2).
374: In Fig.~\ref{fig:DisCon52}.a the square pyramid core structure is
375: connected to two inherent structures, labeled by 2 and 3, by pathways
376: whose energies do not exceed $-13.8$ (in units of $\epsilon_{_{XX}}$).
377: Since isomer \# 2 contains the same core structure
378: (the square pyramid) as the global minimum the corresponding isomerization
379: thus does not lead to a change in the core structure of the cluster. For
380: present purposes, therefore, the barrier that connects them is not a
381: ``relevant'' barrier. The relevant barriers are those that
382: connect inherent structures that contain different core structures.
383: The inherent structure \# 3 contains as the core structure a (distorted)
384: trigonal bipyramid (see Fig.~\ref{fig:coreE52}.1).
385: Therefore, the isomerization barrier that connects the
386: inherent structure \# 3 with global minimum is the lowest relevant
387: isomerization barrier and
388: its value is $\Delta$E$_{1,3}$=0.986$\epsilon_{_{XX}}$.
389: Fig.~\ref{fig:DisCon52}.b and Fig.~\ref{fig:DisCon52}.c show that
390: increasing the value of $\epsilon$ increases isomerization barriers
391: that connect inherent structure \#1 (the square pyramid core structure)
392: with inherent structure \#2 (the distorted trigonal bipyramid core structure).
393: Numerically, these barriers are
394: $\Delta$E$_{1,2}$=1.227$\epsilon_{_{XX}}$ and
395: $\Delta$E$_{1,2}$=1.431$\epsilon_{_{XX}}$, respectively.
396: In Fig.~\ref{fig:DisCon52}.d the square pyramid core structure is
397: connected by an isomerization barrier of
398: $\Delta$E$_{1,2}$=1.647$\epsilon_{_{XX}}$
399: with two (almost degenerate in energy) distorted trigonal bipyramid
400: core structures.
401: As illustrated in Fig.~\ref{fig:DisCon52} and discussed above,
402: the barriers that determine the isomerization
403: kinetics of these newly induced structures are sensitive to the
404: ($\sigma$,$\epsilon$) values and can thus be at least partially controlled.
405: These two simple results are specific demonstrations of goals 1 and 3
406: stated above.
407:
408: \begin{figure}[!htbp] \centering
409: \begin{tabular}{@{}ccc@{}}
410: \includegraphics[width=2.4cm,clip=true]{fig5.1.ps} &
411: \includegraphics[width=2.4cm,clip=true]{fig5.2.ps} &
412: \includegraphics[width=2.4cm,clip=true]{fig5.3.ps} \\
413: (5.1) & (5.2) & (5.3) \\
414: \includegraphics[width=2.8cm,clip=true]{fig5.4.ps} &
415: \includegraphics[width=3.0cm,clip=true]{fig5.5.ps} & \\
416: (5.4) & (5.5) & \\
417: \includegraphics[width=3.2cm,clip=true]{fig5.6.ps} &
418: \includegraphics[width=3.0cm,clip=true]{fig5.7.ps} & \\
419: (5.6) & (5.7) & \\
420: \end{tabular}
421: \caption{\label{fig:coreE52}
422: Plots of $\mathrm{X_5Y_2}$ structures for selected
423: $\mathrm(\sigma,\epsilon)$ values. The decimal number for each figure denotes
424: the corresponding $\mathrm(\sigma,\epsilon)$ domain in
425: Fig.~\ref{fig:corePES52}.}
426: \end{figure}
427: \begin{figure*}
428: \begin{tabular}{@{}cc@{}}
429: \includegraphics[width=8.5cm,clip=true]{fig_6a.ps} &
430: \includegraphics[width=8.5cm,clip=true]{fig_6b.ps} \\
431: (a) & (b) \\
432: \includegraphics[width=8.5cm,clip=true]{fig_6c.ps} &
433: \includegraphics[width=8.5cm,clip=true]{fig_6d.ps} \\
434: (c) & (d) \\
435: \end{tabular}
436: \caption{\label{fig:DisCon52}
437: Disconnectivity graph for $\mathrm{X_5Y_2}$
438: $\mathrm(\sigma,\epsilon)$ values demonstrating that we can control
439: barriers for the selected inherent structures. The energy scale is in
440: units of $\epsilon_{_{XX}}$. The $\mathrm(\sigma,\epsilon)$ values for
441: panels (a--d) are (0.4,0.5), (0.4,1.0), (0.4,1.5) and (0.4,2.0),
442: respectively.}
443: \end{figure*}
444:
445: As a second illustration, we consider mixed clusters of the type
446: $\mathrm{X_7Y_3}$. This system builds upon a parent, seven-atom, ``magic
447: number'' system known to exhibit a set of four, energetically distinct
448: inherent structures. The core inherent structures and associated energies
449: for the stable $\mathrm X_7$ inherent structures are presented in
450: Fig.~\ref{fig:core7}. Figure~\ref{fig:corePES73}, a
451: $\mathrm(\sigma,\epsilon)$ contour plot of the core-atom potential
452: energies of the lowest total energy $\mathrm{X_7Y_3}$ clusters, again
453: reveals the presence of definite ``core-phases''.
454: \begin{figure}
455: \includegraphics[clip=true,width=8.5cm]{fig_7.font.ps}
456: \caption{\label{fig:corePES73}
457: $\mathrm E_{core}(\sigma,\epsilon)$ for $\mathrm{X_7Y_3}$.
458: Format for the plot is
459: the same as in Fig.~\ref{fig:corePES52}.}
460: \end{figure}
461: As illustrated in
462: Fig.~\ref{fig:coreE73}, some of these regions correspond to various
463: core structures present in the parent $\mathrm X_7$ system while others
464: correspond to new structures not seen in the original, single-component
465: cluster. We can see from Figs.~\ref{fig:corePES73}$-$\ref{fig:DisCon73}
466: \begin{figure}[!htbp] \centering
467: \begin{tabular}{@{}cc@{}}
468: \includegraphics[width=3.4cm,clip=true]{fig8.1.ps} &
469: \includegraphics[width=3.4cm,clip=true]{fig8.2.ps} \\
470: (8.1) & (8.2) \\
471: \includegraphics[width=3.4cm,clip=true]{fig8.3.ps} &
472: \includegraphics[width=2.8cm,clip=true]{fig8.4.ps} \\
473: (8.3) & (8.5) \\
474: \includegraphics[width=3.4cm,clip=true]{fig8.5.ps} &
475: \includegraphics[width=3.6cm,clip=true]{fig8.6.ps} \\
476: (8.6) & (8.8) \\
477: \end{tabular}
478: \caption{\label{fig:coreE73}
479: Plots of selected $\mathrm{X_7Y_3}$ structures for various
480: $\mathrm(\sigma,\epsilon)$ values identified in Fig.~\ref{fig:corePES73}.
481: The number of the structures correspond to the regions labeled in
482: Fig.~\ref{fig:corePES73}. Note that many of the core structures for
483: these systems are not stable energy structures of the bare
484: $\mathrm X_7$ system.}
485: \end{figure}
486: \begin{figure*}
487: \begin{tabular}{@{}cc@{}}
488: \includegraphics[width=8.5cm,clip=true]{fig_9a.70.ps} &
489: \includegraphics[width=8.5cm,clip=true]{fig_9b.70.ps} \\
490: (a) & (b) \\
491: \includegraphics[width=8.5cm,clip=true]{fig_9c.70.ps} &
492: \includegraphics[width=8.5cm,clip=true]{fig_9d.70.ps} \\
493: (c) & (d) \\
494: \end{tabular}
495: \caption{\label{fig:DisCon73}
496: Disconnectivity graph for $\mathrm{X_7Y_3(\sigma,\epsilon)}$
497: values demonstrating that we can control barriers for the selected
498: inherent structures. The energy scale is in units of $\epsilon_{_{XX}}$.
499: Only branches leading to the 70 lowest-energy minima are shown.}
500: \end{figure*}
501: that the impurity Y-atoms provide us with significant control over the
502: relative ordering of the core energies of the parent $\mathrm X_7$
503: system. Specifically, by choosing an appropriate range of
504: $\mathrm(\sigma,\epsilon)$ values, we can generate $\mathrm{X_7Y_3}$
505: clusters in which the lowest (total) energy inherent structure can
506: have core structures that are either pentagonal bipyramid, capped octahedral,
507: or bicapped trigonal bipyramidal in nature. Moreover, since we can
508: manipulate the isomerization barriers in these systems, we can at least
509: partially stabilize clusters that exhibit selected core structures with
510: respect to isomerization. This is illustrated in Fig.~\ref{fig:DisCon73}.
511:
512: Figures~\ref{fig:DisCon73}.a~--~\ref{fig:DisCon73}.d represent the
513: $\mathrm{X_7Y_3}$ cluster at four points in Fig.~\ref{fig:corePES73} with
514: $\mathrm{X_7Y_3(\sigma,\epsilon)}$ coordinates (0.4,0.5), (0.4,1.0),
515: (0.4,1.5) and (0.4,2.0), respectively.
516: The number of inherent structures available to the $\mathrm{X_7Y_3}$
517: cluster varies from more than 800 in Fig.~\ref{fig:DisCon73}.a to 400
518: in Fig.~\ref{fig:DisCon73}.d. Since we are primarily interested in
519: energetically low-lying inherent structures we show only lowest 70
520: inherent structures. The global minimum of each system is labeled by
521: number 1 and contains as a recognizable component the core structure
522: shown in Fig.~\ref{fig:coreE73}.5. We should mention that for a given
523: range of $\sigma$ and $\epsilon$ values, ($\sigma, \epsilon$)$\in$ [0.1,2.0],
524: we have not been able to find a global minimum that would contain as
525: a recognizable component inherent structure \# 4 of the parent $\mathrm{X_7}$
526: cluster (see Fig.~\ref{fig:core7}.d). This is the reason why none of the
527: domains in Fig.~\ref{fig:corePES73} is labeled by number 4.
528: In Fig.~\ref{fig:DisCon73}.a the global minimum, the core structure \# 5
529: (see Fig.~\ref{fig:coreE73}.5), is linked to inherent structure \# 2
530: which contains the (distorted) capped octahedron core structure (see
531: Fig.~\ref{fig:coreE73}.2).
532: The isomerization barrier between them is
533: $\Delta$E$_{1,2}$=0.494$\epsilon_{_{XX}}$.
534: %In Fig.~\ref{fig:DisCon73}.b the inherent structure \#2 contains the same
535: %core structure as the global minimum. The corresponding isomerization between
536: %them does not lead to a change in the core structure of the cluster.
537: %On other hand,
538: %the inherent structure \#3 contains as the core structure a distorted
539: %capped octahedron. Therefore, the lowest relevant isomerization barrier
540: %is between the inherent structure \#3 and the global minimum.
541: %The barrier numerical value is $\Delta$E$_{1,2}$=0.975$\epsilon_{_{XX}}$.
542: Fig.~\ref{fig:DisCon73}.b and Fig.~\ref{fig:DisCon73}.c show that
543: increasing the value of $\epsilon$ increases isomerization barriers,
544: that connect inherent structure \#1 (see core structure in
545: Fig~\ref{fig:coreE73}.5)
546: with inherent structure \#3 and \#2
547: (a distorted capped octahedron core structure), respectively.
548: Numerically, these barriers are
549: $\Delta$E$_{1,3}$=0.975$\epsilon_{_{XX}}$ and
550: $\Delta$E$_{1,2}$=1.136$\epsilon_{_{XX}}$, respectively.
551: The inherent structure \#2 in Fig.~\ref{fig:DisCon73}.b contains the same
552: core structure as the global minimum and, therefore, has not been considered
553: relevant for the isomerization (see explanation above for
554: Fig.~\ref{fig:DisCon52}.a).
555: In Fig.~\ref{fig:DisCon52}.d the core structure \# 5 is
556: connected to two, energetically almost degenerate, inherent structures
557: labeled by 2 and 3, by pathways whose energies do not exceed $-40.5$ (in
558: units of $\epsilon_{_{XX}}$). Similarly to the case of
559: Fig.~\ref{fig:DisCon52}.a, the isomer \# 2 contains the same core structure
560: as the global minimum so the corresponding isomerization does not lead
561: to a change in the core structure of the cluster. The barrier that links
562: them is not a relevant barrier. The inherent structure \# 3 contains
563: as the core structure a distorted capped octahedron. Therefore, the
564: isomerization barrier which connects the inherent structure \# 3 with
565: global minimum is the lowest relevant isomerization barrier and its value
566: is $\Delta$E$_{1,3}$=1.304$\epsilon_{_{XX}}$.
567:
568:
569: \section{Conclusions} \label{sec:conclude}
570:
571: In the present work, we have considered the general task of
572: altering core cluster structures. We are, in effect, attempting to turn the
573: logic of the minimization problem upside down. Rather than seeking the
574: global minimum of complex potential energy surfaces, we are instead attempting
575: to exploit what has been learned about the general minimization problem to
576: controllably alter core cluster structures. Specifically, we are examining
577: the extent to which we can induce new core geometries as well as reorder and
578: stabilize existing, higher-lying, local core structures.
579:
580: Our approach, in the present discussion, has been thermodynamic in nature. We
581: have utilized selected adatoms to effect our desired core cluster
582: modifications. We have presented results for two simple binary cluster
583: examples, the $\mathrm{X_5Y_2}$ and $\mathrm{X_7Y_3}$ systems, to validate
584: our approach.
585:
586: We speculate that there are at least two important directions for future
587: theoretical development of the present ideas. One direction will be to
588: explore the use of more complex adsorbates to achieve selected core cluster
589: structures. One could, for example, imagine using ``exterior'' methods in
590: which
591: encapsulating agents of well-defined geometries were utilized to induce desired
592: core structures. Alternatively, ``interior'' approaches in which complex
593: objects, perhaps even previously engineered clusters, could be utilized as
594: ``seeds'' or ``templates'' to produce a desired structure in the
595: surrounding cluster
596: (either globally or locally). Another important direction will be to explore
597: the extent to which previously engineered cluster structures can be assembled
598: using ``cluster assembled materials'' methods to produce larger scale,
599: macroscopic structures. If this proves possible, it would seem to offer an
600: important direction in the production of novel materials starting from
601: synthetic precursors whose core structures and properties are highly varied
602: and are under user control.
603:
604:
605: {\bf Acknowledgment}
606:
607: The authors acknowledge support from the National Science Foundation
608: through awards No. CHE-0095053. and CHE-0131114. They would also like
609: to thank Dr. M. Miller for helpful discussions and for his gracious
610: assistance with respect to the preparation of the disconnectivity graphs
611: in the present paper.
612:
613: \bibliography{sabo}
614:
615: %\pagebreak
616:
617: \end{document}
618: